應用Cascade學習法預測B細胞抗原表位

Abstract

當外來病原體入侵人體時，免疫系統中的B細胞會辨識其抗原表位，並製造抗體與該抗原進行結合，並觸發免疫反應將其消滅，同時，部分B細胞將活化成記憶細胞，並對此抗原產生記憶，若再次遇到相同抗原時，便能加速免疫反應。因此，若能透過電腦計算出正確的抗原表位位置，對於免疫系統研究以及疫苗的研發皆能提供極大的幫助。 目前，已有許多工具被開發來進行B細胞構象或線性表位預測，所使用的方法有藉由胺基酸物理化學性質或是以蛋白質序列作為特徵值，並配合不同的機器學習演算法進行預測，本研究中應用cascade學習法進行B細胞抗原表位預測，經由多種B細胞抗原表位預測工具並配合歸納學習演算法預測抗原表位，而實驗顯示我們所提出的方法皆比使用單獨一種預測器的效能更為優異。

When a pathogen invades the human body, the B cells in the immune system recognize and bind to the pathogen before producing the antibodies against the antigen. The region of an antigen that the B-cell binds to is called an epitope. Because part of the B cells transform into the memory cells which can be reactivated efficiently on a second encounter with the same pathogen, an accurate computational method for epitopic region prediction can accelerate the research in immune systems and vaccine designs. Many tools have been developed to predict conformational or linear epitopes. They use different physicochemical properties as features, and adopt various computational models. In this study, we applied cascade learning to B-cell epitope prediction. It learned from multiple B-cell epitope predictors by cascading various inductive classifiers to produce the final prediction. The experimental results show that the performance of the proposed method is superior to those of the existing prediction tools.